A two-layer nested heterogeneous ensemble learning predictive method for COVID-19 mortality

Shaoze Cui, Yanzhang Wang, Dujuan Wang, Qian Sai, Ziheng Huang, T. C.E. Cheng

Research output: Journal article publicationJournal articleAcademic researchpeer-review

1 Citation (Scopus)

Abstract

The COVID-19 epidemic has had a great adverse impact on the world, having taken a heavy toll, killing hundreds of thousands of people. In order to help the world better combat COVID-19 and reduce its death toll, this study focuses on the COVID-19 mortality. First, using the multiple stepwise regression analysis method, the factors from eight aspects (economy, society, climate etc.) that may affect the mortality rates of COVID-19 in various countries is examined. In addition, a two-layer nested heterogeneous ensemble learning-based prediction method that combines linear regression (LR), support vector machine (SVM), and extreme learning machine (ELM) is developed to predict the development trends of COVID-19 mortality in various countries. Based on data from 79 countries, the experiment proves that age structure (proportion of the population over 70 years old) and medical resources (number of beds) are the main factors affecting the mortality of COVID-19 in each country. In addition, it is found that the number of nucleic acid tests and climatic factors are correlated with COVID-19 mortality. At the same time, when predicting COVID-19 mortality, the proposed heterogeneous ensemble learning-based prediction method shows better prediction ability than state-of-the-art machine learning methods such as LR, SVM, ELM, random forest (RF), long short-term memory (LSTM) etc.

Original languageEnglish
Article number107946
JournalApplied Soft Computing
Volume113
DOIs
Publication statusPublished - Dec 2021

Keywords

  • COVID-19
  • Ensemble learning
  • Hybrid method
  • Mortality
  • Stepwise multiple regression
  • Time series prediction

ASJC Scopus subject areas

  • Software

Cite this